ROS Rankings
Every player's value for the rest of the redraft season. Projected points above a waiver pickup, moved by what real leagues pay in trades. The trade calculator uses the same numbers.
How the ROS rankings are built
The projection. Two models trained on every NFL season since 2014. One predicts a player's points per game when he plays, from his age, experience, draft slot, last season, this season, his share of targets and carries, team and quarterback changes, and his offense's betting lines. The other predicts the chance he plays each remaining week, from his roster and injury status, the injury, and how long players in that spot have actually stayed out. His next game's prop lines then move the rate, weighted by how much lines have predicted the rest of the season before. When the news gives a return timetable, that moves the weekly chances too.
The value. For every week he has left, the chance he plays times the points he scores above the best player a 12-team league leaves on waivers. An injured star keeps the weeks he's expected back. This is the Projection column.
How good it is. Tested on the 2024 and 2025 seasons, which it never trained on, it missed rest-of-season points by 10% less than our previous projection and ranked players noticeably closer to how they finished. The details are in the footer.
The trades. A crawler reads public Sleeper redraft leagues and records every completed two-team trade: the players each side got, plus any FAAB. A side is worth the sum of its players. Leagues only complete trades both managers accept, so we move each player's value until the trades people made come out as even as possible. Recent trades count more (two-week half-life).
Why the projection sets the scale. An accepted trade only tells you two sides were about even. It never tells you one player is worth more than another. Fit on trades alone, every value slides toward the same number. So the projection decides how far apart the top and bottom are, and trades decide the order. How many lesser players one star is worth is checked against 2-for-1 trades the fit never saw. A heavily traded player is set mostly by his trades. A rarely traded one stays close to his projection.
Buy low, sell high. For each player traded at least ten times, we compare his ROS value with the value the projection alone would rank him at. A gap over 30% gets a tag.
Dynasty. Not included. Dynasty trades price future seasons, which a rest-of-season ranking shouldn't.